The issue we are running into with the couchbase spark connector under databricks is that there seem to be classpath issues which based on my analysis and limited understandung of the underlying runtime should not happen.
I'm sharing the rough details below, if someone needs access to couchbase cloud credentials to try and reproduce please just ping me directly @daschl on twitter and I'm more than happy to provide those.
The Connector uses the Couchbase Scala SDK underneath which has a direct dependency on project reactor and makes use of it quite heavily.
Looking at the classpath, it looks like there is more than one reactor version there: